Dataset Input/Output Configuration


I have a dataset with around 1000 variables. How do I configure the CSV file such that all but the last variable are configured as InputPort0, with the last being OutputPort0? Deep Learning Studio (Desktop) isn’t doing it automatically, which I know it does under certain conditions.



Hi Jordan,

When uploading dataset, you can include a file named meta.json (along with train.csv). This meta.json can contain information about which columns belong to InputPort0 and which belong to OutputPort0.

Here is example of meta.json for MNIST dataset.

{“Image”: {“type”: “Image”, “port”: “InputPort0”, “categories”: 70000}, “Digit Label”: {“type”: “Categorical”, “port”: “OutputPort0”, “categories”: 10}}


1 Like

Excellent. Thank you, Rajendra!

  • Jordan

Hi, must i create the json? or it is created by deep learning studio when the train csv is uploaded?


Same question i tryed to make the meta.json and zipped it with the dataset but no luck.

{“Input1”: {“type”: “Image”, “port”: “InputPort0”, “categories”: 18},“Input2”: {“type”: “Image”, “port”: “InputPort1”, “categories”: 18}, “Result”: {“type”: “Image”, “port”: “OutputPort0”, “categories”: 1}}

Hi in meta.json file port name should be InputPort0 for all the input data fields.